EPTMetalopolis prone to close positions, why? - Page 10
| Forum Index > SC2 General |
I apologize to everyone in this thread for taking the OP seriously. My mod senses are definitely off today. -- Chill | ||
|
Centorian
United States95 Posts
| ||
|
Myia
173 Posts
| ||
|
Ihle
Norway36 Posts
On February 16 2011 07:06 gerundium wrote: He could definitely have gotten the 1:10000 shot, it its at least way more likely that we are reading about a 1000 to 1 shot on here, because posting about something like this becomes more likely as someone tests this and comes up with a statistical outlier. It is not a statistical outlier, if everyone in the world opened metalopolis all day they would not get 72 out of 100, given the probability is 1/3. This is either a troll or a fault on the op, or it is a fault with the game. | ||
Zelniq
United States7166 Posts
On February 16 2011 06:40 TheGrimace wrote: ... How to check in sc2gears: 1. Open sc2gears, go to the replay analyzer and set chart type to "map view" 2. Drag and drop each replay 3. Note the spawn positions 4. Repeat ... I used this method and went through 50 of my metalopolis games in my autosaved replay folder (dating from Jan 13 to Jan 24th) and found 15 games where I spawned close positions. I didn't bother to tally if the others were either close by air or cross, tho I probably should have. 15 / 50 = .3, 30% of the time spawning close spots. when I have more time later I'll go through more but just putting this out there for now. I suppose you'll just have to take my word for it I forgot to check if any of these were vs the AI, but i'm pretty sure almost all or probably all were 1v1 player vs player, cus my test games vs the ai were usually on other maps, closer to the top of the map list  | ||
|
Deadeight
United Kingdom1629 Posts
On February 16 2011 07:07 aristarchus wrote: 70% is far enough from 1/3 to be [...] definitely obvious to a pro who is carefully practicing the map. My thoughts exactly. Kind of odd that the OP hasn't been posting too. | ||
|
annul
United States2841 Posts
| ||
|
Pyrrhuloxia
United States6700 Posts
+ Show Spoiler + All tests were done as Terran vs. Random AI. Player 1 was me as orange and Player 2 was a Medium AI as blue. All tests were 1v1, Melee, Faster, and Locked Alliances. 31.25% Close Positions 10/32 43.75% (Close by) Air Positions 14/32 25% Far Positions 8/32 Player 1 Spawn Locations: 9/32 Left 9/32 Down 8/32 Right 6/32 Up Player 2 Spawn Locations: 11/32 Up 8/32 Left 10/32 Down 3/32 Right Player 2's Race (set to Random) 6/32 Zerg 13/32 Terran 13/32 Protoss Hypothesis: Randomizer is working fine. P values are as overvalued as I perceived when I was in Statistics class, and lead to all sorts of great scientific discoveries like ESP: http://www.nytimes.com/2011/01/11/science/11esp.html After 64: Close Positions: 18/64 Air Positions: 24/64 Far Positions: 24/64 Player 1 Spawn Locations 17/64 Left 13/64 Down 16/64 Right 18/64 Up Player 2 Spawn Locations 15/64 Up 12/64 Left 26/64 Down 11/64 Right Player 2's Race 16/64 Zerg 22/64 Terran 25/64 Protoss | ||
|
deejY
Germany44 Posts
9x cross 13x close-air 8x close | ||
|
Dimagus
United States1004 Posts
| ||
|
TheGrimace
United States929 Posts
Close: 9 Air: 9 Cross: 13 So 29% Close and Air, 42% Cross. I'll be interested to see the results of a single person performing the task in a smallish time frame. Keep it up Pyrrhuloxia. | ||
|
eighteen8
105 Posts
close gnd position: 37,2% close air position: 37,2% cross position: 25,6% on a total number of 45 replays for meta how about all these replay websites - there are some kind of statistics on their page (i guess these are values from uploaded replays)...maybe its also possible to do this with spawn positions. | ||
|
Souai
United States47 Posts
Results were: 37 Close 34 Air 29 Cross It seems the OP had a very unusual experience, used a tainted data pool, or is trolling everyone, shrug. | ||
|
pelicanguy
1 Post
Using an exact binomial test with p = 1/3 and number of successes = 72: P-value = 3.338*10^-15 95% confidence interval: (.62,.81) In case you don't understand what these numbers mean, there is a 3.338*10^-17 percentage chance that you will get 72 or more close position spawns or an equivalently small number of close spawns or less out of 100 games with the probability of a close spawn being 1/3. The confidence interval is an estimate of what the actual probability of spawning in close positions is, so it's about 62% to 81%. Most likely, this is just a result of pseudo-random number generators failing their job. A larger sample size will always be helpful, but from the OP's data alone you can see that the probability of a "fair" spawner is astronomically low. In fact, in order to expect to see such strange data or stranger if the probability of close spawns was actually 1/3, you would have to do about 3*10^16 trials of 100 games. That's 3000000000000000000 games of starcraft 2. | ||
|
Ludwigvan
Germany2371 Posts
 Really, OP, tell us exactly how you did this. Your test method was most likely wrong. What was your battle net name again, Hwang Woo-suk? | ||
|
ShadowDrgn
United States2497 Posts
On February 16 2011 06:56 Chill wrote: Right, which is why I'm suggesting a tournament replay pack, such as the TLOpen or similar. I just went through the MLG Dallas replay pack I had lying around in SC2gears: Close ground: 40 Close air: 30 Cross: 32 12.5% of having 40/102 or more in close ground spots so that's not unreasonable. Definitely nothing like the OP. | ||
|
Fraidnot
United States824 Posts
| ||
|
EcterA
United States949 Posts
Close Ground - 10 - 32.26% Close Air - 7 - 22.58% Cross - 14 - 45.16% I was going to do more, but it really is pretty boring, and it seems like a general consensus in the thread is that the OP's results might have been an anomoly, and that there is a better way to test it. I'll do more if the OP ever comes back to the discussion. | ||
| ||
